feat: 更新文档链接，增强客户端接入指南，优化代理日志记录和错误处理
This commit is contained in:
ciat-777
@@ -39,12 +39,12 @@
|
|||||||
</p>
|
</p>
|
||||||
|
|
||||||
<p align="center">
|
<p align="center">
|
||||||
<a href="docs/index.md"><strong>📚 文档中心</strong></a> ·
|
<a href="https://metapi.cita777.me"><strong>📚 在线文档</strong></a> ·
|
||||||
<a href="docs/getting-started.md">快速上手</a> ·
|
<a href="https://metapi.cita777.me/getting-started">快速上手</a> ·
|
||||||
<a href="docs/deployment.md">部署指南</a> ·
|
<a href="https://metapi.cita777.me/deployment">部署指南</a> ·
|
||||||
<a href="docs/configuration.md">配置说明</a> ·
|
<a href="https://metapi.cita777.me/configuration">配置说明</a> ·
|
||||||
<a href="docs/faq.md">常见问题</a> ·
|
<a href="https://metapi.cita777.me/client-integration">客户端接入</a> ·
|
||||||
<a href="CONTRIBUTING.md">贡献指南</a>
|
<a href="https://metapi.cita777.me/faq">常见问题</a>
|
||||||
</p>
|
</p>
|
||||||
|
|
||||||
</div>
|
</div>
|
||||||
@@ -236,56 +236,18 @@ graph LR
|
|||||||
<p><sub>模型广场 — 一站式浏览所有可用模型的覆盖率、定价和性能指标</sub></p>
|
<p><sub>模型广场 — 一站式浏览所有可用模型的覆盖率、定价和性能指标</sub></p>
|
||||||
</div>
|
</div>
|
||||||
|
|
||||||
### ✅ 自动签到
|
### ✅ 自动签到 · 💰 余额管理 · 🔔 告警通知 · 📊 数据看板
|
||||||
|
|
||||||
- Cron 定时自动签到（默认每天 08:00）
|
- **自动签到**：Cron 定时执行，智能解析奖励金额，签到失败自动通知
|
||||||
- 智能解析签到奖励金额，签到失败自动通知
|
- **余额管理**：定时刷新，收入追踪，余额兜底估算，凭证过期自动重登录
|
||||||
- 逐账号独立执行，支持启用/禁用控制
|
- **告警通知**：支持 Webhook / Bark / Server酱 / SMTP 邮件，覆盖余额不足、站点异常、签到失败等场景
|
||||||
- 完整的签到日志记录，支持历史查询
|
- **数据看板**：余额分布饼图、消费趋势图表、全局搜索、代理请求日志
|
||||||
- 并发锁机制，防止重复签到
|
|
||||||
|
|
||||||
### 💰 余额管理
|
|
||||||
|
|
||||||
- 定时余额刷新（默认每小时），批量更新所有活跃账号
|
|
||||||
- 收入追踪：记录每日/累计收入，追踪额度消耗趋势
|
|
||||||
- 余额兜底估算：API 不可用时，从代理日志推算余额变动
|
|
||||||
- 自动重登录：凭证过期时自动刷新
|
|
||||||
|
|
||||||
### 🔔 告警与通知
|
|
||||||
|
|
||||||
支持四种通知渠道：
|
|
||||||
|
|
||||||
| 渠道 | 说明 |
|
|
||||||
| ------------------- | -------------------- |
|
|
||||||
| **Webhook** | 自定义 HTTP 推送 |
|
|
||||||
| **Bark** | iOS 推送通知 |
|
|
||||||
| **Server酱** | 微信 / Telegram 通知 |
|
|
||||||
| **SMTP 邮件** | 标准邮件通知 |
|
|
||||||
|
|
||||||
告警场景覆盖：余额不足预警、站点/账号异常、签到失败、代理请求失败、Token 过期提醒、每日摘要报告。支持告警冷却机制（默认 300 秒），防止相同事件重复通知。
|
|
||||||
|
|
||||||
### 📊 数据看板
|
|
||||||
|
|
||||||
- 站点余额分布饼图、每日消费趋势图表
|
|
||||||
- 全局搜索（站点、账号、模型）
|
|
||||||
- 系统事件日志、代理请求日志（含模型、状态、延迟、Token 消耗、成本估算）
|
|
||||||
|
|
||||||
<div align="center">
|
<div align="center">
|
||||||
<img src="docs/screenshots/dashboard.png" alt="dashboard-detail" width="700"/>
|
<img src="docs/screenshots/dashboard.png" alt="dashboard-detail" width="700"/>
|
||||||
<p><sub>数据看板 — 余额分布、消费趋势、系统健康状态一目了然</sub></p>
|
<p><sub>数据看板 — 余额分布、消费趋势、系统健康状态一目了然</sub></p>
|
||||||
</div>
|
</div>
|
||||||
|
|
||||||
### 🔬 模型操练场
|
|
||||||
|
|
||||||
- 在线交互式对话测试，即时验证模型可用性与响应质量
|
|
||||||
- 支持选择任意已路由模型，对比不同通道的输出效果
|
|
||||||
- 流式 / 非流式双模式测试
|
|
||||||
|
|
||||||
<div align="center">
|
|
||||||
<img src="docs/screenshots/playground.png" alt="playground-detail" width="700"/>
|
|
||||||
<p><sub>模型操练场 — 在线交互式测试，验证模型可用性与响应质量</sub></p>
|
|
||||||
</div>
|
|
||||||
|
|
||||||
### 📦 轻量部署
|
### 📦 轻量部署
|
||||||
|
|
||||||
- **单 Docker 容器**，内置 SQLite，无外部依赖
|
- **单 Docker 容器**，内置 SQLite，无外部依赖
|
||||||
@@ -296,336 +258,28 @@ graph LR
|
|||||||
|
|
||||||
## 🚀 快速开始
|
## 🚀 快速开始
|
||||||
|
|
||||||
### Zeabur 一键部署
|
|
||||||
|
|
||||||
<a href="https://zeabur.com/templates/DOX5PR">
|
<a href="https://zeabur.com/templates/DOX5PR">
|
||||||
<img alt="Deploy on Zeabur" src="https://zeabur.com/button.svg" height="28">
|
<img alt="Deploy on Zeabur" src="https://zeabur.com/button.svg" height="28">
|
||||||
</a>
|
</a>
|
||||||
|
|
||||||
点击按钮即可一键部署到 [Zeabur](https://zeabur.com)，自动配置容器、持久化存储和域名，部署后按提示填写 `AUTH_TOKEN` 和 `PROXY_TOKEN` 即可使用。
|
|
||||||
|
|
||||||
### Docker Compose（推荐）
|
|
||||||
|
|
||||||
```bash
|
```bash
|
||||||
mkdir metapi && cd metapi
|
docker run -d --name metapi -p 4000:4000 \
|
||||||
|
|
||||||
cat > docker-compose.yml << 'EOF'
|
|
||||||
services:
|
|
||||||
metapi:
|
|
||||||
image: 1467078763/metapi:latest
|
|
||||||
ports:
|
|
||||||
- "4000:4000"
|
|
||||||
volumes:
|
|
||||||
- ./data:/app/data
|
|
||||||
environment:
|
|
||||||
AUTH_TOKEN: ${AUTH_TOKEN:?AUTH_TOKEN is required}
|
|
||||||
PROXY_TOKEN: ${PROXY_TOKEN:?PROXY_TOKEN is required}
|
|
||||||
CHECKIN_CRON: "0 8 * * *"
|
|
||||||
BALANCE_REFRESH_CRON: "0 * * * *"
|
|
||||||
PORT: ${PORT:-4000}
|
|
||||||
DATA_DIR: /app/data
|
|
||||||
TZ: ${TZ:-Asia/Shanghai}
|
|
||||||
restart: unless-stopped
|
|
||||||
EOF
|
|
||||||
|
|
||||||
# 设置令牌并启动
|
|
||||||
# AUTH_TOKEN = 管理后台初始管理员令牌（登录后台时输入这个值）
|
|
||||||
export AUTH_TOKEN=your-admin-token
|
|
||||||
# PROXY_TOKEN = 下游客户端调用 /v1/* 使用的令牌
|
|
||||||
export PROXY_TOKEN=your-proxy-sk-token
|
|
||||||
docker compose up -d
|
|
||||||
```
|
|
||||||
|
|
||||||
<details>
|
|
||||||
<summary><strong>Docker 命令一行启动</strong></summary>
|
|
||||||
|
|
||||||
```bash
|
|
||||||
docker run -d --name metapi \
|
|
||||||
-p 4000:4000 \
|
|
||||||
-e AUTH_TOKEN=your-admin-token \
|
-e AUTH_TOKEN=your-admin-token \
|
||||||
-e PROXY_TOKEN=your-proxy-sk-token \
|
-e PROXY_TOKEN=your-proxy-sk-token \
|
||||||
-e TZ=Asia/Shanghai \
|
-v ./data:/app/data --restart unless-stopped \
|
||||||
-v ./data:/app/data \
|
|
||||||
--restart unless-stopped \
|
|
||||||
1467078763/metapi:latest
|
1467078763/metapi:latest
|
||||||
```
|
```
|
||||||
|
|
||||||
</details>
|
启动后访问 `http://localhost:4000`，用 `AUTH_TOKEN` 登录即可。
|
||||||
|
|
||||||
🎉 启动后访问 `http://localhost:4000`，用 `AUTH_TOKEN` 登录即可！
|
|
||||||
|
|
||||||
> [!IMPORTANT]
|
> [!IMPORTANT]
|
||||||
> 请务必修改 `AUTH_TOKEN` 和 `PROXY_TOKEN`，不要使用默认值。数据存储在 `./data` 目录中，升级不影响已有数据。
|
> 请务必修改 `AUTH_TOKEN` 和 `PROXY_TOKEN`，不要使用默认值。
|
||||||
|
|
||||||
> [!TIP]
|
📖 **[完整部署文档](https://metapi.cita777.me/deployment)** — Zeabur 一键部署 / Docker Compose / Release 包 / 反向代理 / 升级回滚
|
||||||
> 初始管理员令牌就是启动时配置的 `AUTH_TOKEN`。
|
|
||||||
> 如果是非 Compose 场景且你没有显式设置 `AUTH_TOKEN`，默认值为 `change-me-admin-token`（仅用于本地调试）。
|
|
||||||
> 若你在后台「设置」里修改过管理员令牌，后续登录请使用修改后的新令牌。
|
|
||||||
|
|
||||||
### Release 包启动（Linux / macOS / Windows）
|
📖 **[环境变量与配置](https://metapi.cita777.me/configuration)** — 全部环境变量、路由参数、通知渠道
|
||||||
|
|
||||||
如果你不想用 Docker，也可以直接下载预打包的 Release 产物运行：
|
📖 **[客户端接入指南](https://metapi.cita777.me/client-integration)** — Cursor / Claude Code / Codex / Open WebUI 等配置方法
|
||||||
|
|
||||||
1. 打开 [Releases](https://github.com/cita-777/metapi/releases) 下载与你系统匹配的压缩包
|
|
||||||
2. 解压后进入目录
|
|
||||||
3. 设置环境变量并启动
|
|
||||||
|
|
||||||
Linux / macOS：
|
|
||||||
|
|
||||||
```bash
|
|
||||||
export AUTH_TOKEN=your-admin-token
|
|
||||||
export PROXY_TOKEN=your-proxy-sk-token
|
|
||||||
export PORT=4000
|
|
||||||
export DATA_DIR=./data
|
|
||||||
./start.sh
|
|
||||||
```
|
|
||||||
|
|
||||||
Windows（PowerShell）：
|
|
||||||
|
|
||||||
```powershell
|
|
||||||
$env:AUTH_TOKEN="your-admin-token"
|
|
||||||
$env:PROXY_TOKEN="your-proxy-sk-token"
|
|
||||||
$env:PORT="4000"
|
|
||||||
$env:DATA_DIR="./data"
|
|
||||||
.\start.bat
|
|
||||||
```
|
|
||||||
|
|
||||||
`start.sh` / `start.bat` 的作用是：
|
|
||||||
|
|
||||||
- 先检查 `better-sqlite3` 与当前 Node.js ABI 是否兼容
|
|
||||||
- 若检测到 ABI 不匹配，会自动尝试 `npm rebuild better-sqlite3`
|
|
||||||
- 若重建失败，会回退到 `npm ci --omit=dev` 重新安装运行时依赖
|
|
||||||
- 最后执行数据库迁移并启动服务
|
|
||||||
|
|
||||||
> [!NOTE]
|
|
||||||
> Release 包依赖本机安装 Node.js（支持 20+，推荐 22 LTS）。
|
|
||||||
> 如果本机 Node 主版本与打包时不同（例如包内依赖基于 Node 22，而本机是 Node 24），首次启动可能触发自动重建，需联网。
|
|
||||||
|
|
||||||
### 升级
|
|
||||||
|
|
||||||
```bash
|
|
||||||
docker compose pull && docker compose up -d && docker image prune -f
|
|
||||||
```
|
|
||||||
|
|
||||||
📖 更详细的部署方式请参考 [部署指南](docs/deployment.md)
|
|
||||||
|
|
||||||
---
|
|
||||||
|
|
||||||
## 📚 文档中心
|
|
||||||
|
|
||||||
> 文档站（VitePress）本地预览：
|
|
||||||
>
|
|
||||||
> ```bash
|
|
||||||
> npm run docs:dev
|
|
||||||
> ```
|
|
||||||
>
|
|
||||||
> 文档站构建：
|
|
||||||
>
|
|
||||||
> ```bash
|
|
||||||
> npm run docs:build
|
|
||||||
> ```
|
|
||||||
>
|
|
||||||
> GitHub Actions 自动发布：推送到 `main` 分支后，会执行 `.github/workflows/docs-pages.yml` 并部署到 GitHub Pages。
|
|
||||||
> 首次使用需在仓库设置中开启：
|
|
||||||
> `Settings -> Pages -> Build and deployment -> Source: GitHub Actions`
|
|
||||||
|
|
||||||
| 分类 | 链接 | 说明 |
|
|
||||||
| ------------- | ----------------------------------------------------- | -------------------------------------- |
|
|
||||||
| 🏠 文档站首页 | [docs/index.md](docs/index.md) | 可发布文档站入口（导航/侧栏/搜索） |
|
|
||||||
| 📖 文档总览 | [docs/README.md](docs/README.md) | 文档导航与索引 |
|
|
||||||
| 🚀 快速上手 | [docs/getting-started.md](docs/getting-started.md) | 10 分钟启动 |
|
|
||||||
| 🚢 部署指南 | [docs/deployment.md](docs/deployment.md) | Docker Compose、反向代理、升级回滚 |
|
|
||||||
| ⚙️ 配置说明 | [docs/configuration.md](docs/configuration.md) | 全部环境变量与路由参数 |
|
|
||||||
| 🔌 客户端接入 | [docs/client-integration.md](docs/client-integration.md) | Open WebUI / Cherry Studio / Cursor 等 |
|
|
||||||
| 🔧 运维手册 | [docs/operations.md](docs/operations.md) | 备份恢复、日志排查、健康检查 |
|
|
||||||
| ❓ 常见问题 | [docs/faq.md](docs/faq.md) | 常见报错与修复路径 |
|
|
||||||
| 🧩 经验沉淀规范 | [docs/community/faq-tutorial-guidelines.md](docs/community/faq-tutorial-guidelines.md) | FAQ 与教程提交流程、模板与命名规则 |
|
|
||||||
|
|
||||||
---
|
|
||||||
|
|
||||||
## ⚙️ 环境变量
|
|
||||||
|
|
||||||
### 基础配置
|
|
||||||
|
|
||||||
| 变量名 | 说明 | 默认值 |
|
|
||||||
| ----------------------------- | ------------------------------------------- | ----------------------- |
|
|
||||||
| `AUTH_TOKEN` | 管理后台登录令牌（**必须修改**） | `change-me-admin-token` |
|
|
||||||
| `PROXY_TOKEN` | 代理 API Bearer Token（**必须修改**） | `change-me-proxy-sk-token` |
|
|
||||||
| `PORT` | 服务监听端口 | `4000` |
|
|
||||||
| `DATA_DIR` | 数据目录（SQLite 数据库） | `./data` |
|
|
||||||
| `TZ` | 时区 | `Asia/Shanghai` |
|
|
||||||
| `ACCOUNT_CREDENTIAL_SECRET` | 账号凭证加密密钥 | 默认使用 `AUTH_TOKEN` |
|
|
||||||
|
|
||||||
### 定时任务
|
|
||||||
|
|
||||||
| 变量名 | 说明 | 默认值 |
|
|
||||||
| ------------------------ | -------------------- | ------------- |
|
|
||||||
| `CHECKIN_CRON` | 自动签到 Cron 表达式 | `0 8 * * *` |
|
|
||||||
| `BALANCE_REFRESH_CRON` | 余额刷新 Cron 表达式 | `0 * * * *` |
|
|
||||||
|
|
||||||
<details>
|
|
||||||
<summary><strong>智能路由参数</strong></summary>
|
|
||||||
|
|
||||||
| 变量名 | 说明 | 默认值 |
|
|
||||||
| ------------------------------ | ---------------------- | ------- |
|
|
||||||
| `ROUTING_FALLBACK_UNIT_COST` | 无成本信号时的默认单价 | `1` |
|
|
||||||
| `BASE_WEIGHT_FACTOR` | 基础权重因子 | `0.5` |
|
|
||||||
| `VALUE_SCORE_FACTOR` | 性价比评分因子 | `0.5` |
|
|
||||||
| `COST_WEIGHT` | 路由选择中成本权重 | `0.4` |
|
|
||||||
| `BALANCE_WEIGHT` | 路由选择中余额权重 | `0.3` |
|
|
||||||
| `USAGE_WEIGHT` | 路由选择中使用率权重 | `0.3` |
|
|
||||||
|
|
||||||
</details>
|
|
||||||
|
|
||||||
<details>
|
|
||||||
<summary><strong>通知渠道配置</strong></summary>
|
|
||||||
|
|
||||||
| 变量名 | 说明 | 默认值 |
|
|
||||||
| ----------------------------- | ---------------------- | --------- |
|
|
||||||
| `WEBHOOK_ENABLED` | 启用 Webhook 通知 | `true` |
|
|
||||||
| `WEBHOOK_URL` | Webhook 推送地址 | 空 |
|
|
||||||
| `BARK_ENABLED` | 启用 Bark 推送 | `true` |
|
|
||||||
| `BARK_URL` | Bark 推送地址 | 空 |
|
|
||||||
| `SERVERCHAN_ENABLED` | 启用 Server酱 通知 | `true` |
|
|
||||||
| `SERVERCHAN_KEY` | Server酱 SendKey | 空 |
|
|
||||||
| `SMTP_ENABLED` | 启用邮件通知 | `false` |
|
|
||||||
| `SMTP_HOST` | SMTP 服务器地址 | 空 |
|
|
||||||
| `SMTP_PORT` | SMTP 端口 | `587` |
|
|
||||||
| `SMTP_SECURE` | 使用 SSL/TLS | `false` |
|
|
||||||
| `SMTP_USER` / `SMTP_PASS` | SMTP 认证 | 空 |
|
|
||||||
| `SMTP_FROM` / `SMTP_TO` | 发件/收件人 | 空 |
|
|
||||||
| `NOTIFY_COOLDOWN_SEC` | 相同告警冷却时间（秒） | `300` |
|
|
||||||
|
|
||||||
</details>
|
|
||||||
|
|
||||||
<details>
|
|
||||||
<summary><strong>安全配置</strong></summary>
|
|
||||||
|
|
||||||
| 变量名 | 说明 | 默认值 |
|
|
||||||
| ---------------------- | ---------------------------- | ------------ |
|
|
||||||
| `ADMIN_IP_ALLOWLIST` | 管理端 IP 白名单（逗号分隔） | 空（不限制） |
|
|
||||||
|
|
||||||
</details>
|
|
||||||
|
|
||||||
📖 完整配置说明：[docs/configuration.md](docs/configuration.md)
|
|
||||||
|
|
||||||
---
|
|
||||||
|
|
||||||
## 📡 代理接口
|
|
||||||
|
|
||||||
Metapi 对下游暴露标准 OpenAI / Claude 兼容接口：
|
|
||||||
|
|
||||||
| 接口 | 方法 | 说明 |
|
|
||||||
| -------------------------- | ---- | ---------------------------- |
|
|
||||||
| `/v1/responses` | POST | OpenAI Responses |
|
|
||||||
| `/v1/chat/completions` | POST | OpenAI Chat Completions |
|
|
||||||
| `/v1/messages` | POST | Claude Messages |
|
|
||||||
| `/v1/completions` | POST | OpenAI Completions（Legacy） |
|
|
||||||
| `/v1/embeddings` | POST | 向量嵌入 |
|
|
||||||
| `/v1/images/generations` | POST | 图像生成 |
|
|
||||||
| `/v1/models` | GET | 获取所有可用模型列表 |
|
|
||||||
|
|
||||||
请求头携带 `Authorization: Bearer <PROXY_TOKEN>` 即可访问。
|
|
||||||
|
|
||||||
默认可使用全局 `PROXY_TOKEN`。
|
|
||||||
从 `系统设置 -> 下游 API Key 策略` 可创建多个项目级下游 Key，并单独配置：
|
|
||||||
|
|
||||||
- 过期时间（ExpireAt）
|
|
||||||
- 费用上限与请求上限（MaxCost / MaxRequests）
|
|
||||||
- 模型白名单（SupportedModels，支持 exact/glob/re:regex）
|
|
||||||
- 路由白名单（AllowedRouteIds）
|
|
||||||
- 站点倍率（SiteWeightMultipliers，用于控制不同项目的上游偏好占比）
|
|
||||||
|
|
||||||
---
|
|
||||||
|
|
||||||
## 🔌 接入下游客户端
|
|
||||||
|
|
||||||
适用于所有兼容 OpenAI API 的客户端：
|
|
||||||
|
|
||||||
| 配置项 | 值 |
|
|
||||||
| ------------------ | ----------------------------------------------------------- |
|
|
||||||
| **Base URL** | `https://your-domain.com`（客户端一般会自动拼接 `/v1`） |
|
|
||||||
| **API Key** | 你设置的 `PROXY_TOKEN` 值 |
|
|
||||||
| **模型列表** | 自动从 `GET /v1/models` 获取 |
|
|
||||||
|
|
||||||
### CLI 快速配置（Claude Code / Codex）
|
|
||||||
|
|
||||||
#### Claude Code（`~/.claude/settings.json`）
|
|
||||||
|
|
||||||
```json
|
|
||||||
{
|
|
||||||
"env": {
|
|
||||||
"ANTHROPIC_BASE_URL": "https://your-domain.com",
|
|
||||||
"ANTHROPIC_API_KEY": "your-proxy-sk-token",
|
|
||||||
"ANTHROPIC_AUTH_TOKEN": "your-proxy-sk-token",
|
|
||||||
"CLAUDE_CODE_ATTRIBUTION_HEADER": "0",
|
|
||||||
"CLAUDE_CODE_DISABLE_NONESSENTIAL_TRAFFIC": "1"
|
|
||||||
}
|
|
||||||
}
|
|
||||||
```
|
|
||||||
|
|
||||||
> 说明：不同版本客户端可能读取 `ANTHROPIC_API_KEY` 或 `ANTHROPIC_AUTH_TOKEN`，建议同时设置。
|
|
||||||
|
|
||||||
#### Codex（`~/.codex/config.toml` + `~/.codex/auth.json`）
|
|
||||||
|
|
||||||
`~/.codex/config.toml`
|
|
||||||
|
|
||||||
```toml
|
|
||||||
model = "gpt-5"
|
|
||||||
model_provider = "metapi"
|
|
||||||
|
|
||||||
[model_providers.metapi]
|
|
||||||
name = "metapi"
|
|
||||||
base_url = "https://your-domain.com/v1"
|
|
||||||
```
|
|
||||||
|
|
||||||
`~/.codex/auth.json`
|
|
||||||
|
|
||||||
```json
|
|
||||||
{
|
|
||||||
"OPENAI_API_KEY": "your-proxy-sk-token"
|
|
||||||
}
|
|
||||||
```
|
|
||||||
|
|
||||||
> 提示：`model` 需要是你在 Metapi `GET /v1/models` 可见的模型名。
|
|
||||||
|
|
||||||
### 已验证兼容的客户端
|
|
||||||
|
|
||||||
- [ChatGPT-Next-Web](https://github.com/ChatGPTNextWeb/ChatGPT-Next-Web)
|
|
||||||
- [Open WebUI](https://github.com/open-webui/open-webui)
|
|
||||||
- [Cherry Studio](https://github.com/kangfenmao/cherry-studio)
|
|
||||||
- [Cursor](https://cursor.sh)
|
|
||||||
- [Claude Code](https://docs.anthropic.com/en/docs/claude-code)
|
|
||||||
- Codex CLI
|
|
||||||
- [Roo Code](https://github.com/RooVetGit/Roo-Code)
|
|
||||||
- [Kilo Code](https://github.com/kilocode/kilocode)
|
|
||||||
- 以及所有支持 OpenAI API 格式的客户端
|
|
||||||
|
|
||||||
<details>
|
|
||||||
<summary><strong>常见问题：流式响应异常</strong></summary>
|
|
||||||
|
|
||||||
如果非流式正常但流式异常，请排查：
|
|
||||||
|
|
||||||
1. 反向代理是否关闭了 SSE 缓冲（Nginx 需设置 `proxy_buffering off`）
|
|
||||||
2. 中间层是否改写了 `text/event-stream` Content-Type
|
|
||||||
3. 客户端是否强制要求特定流式格式
|
|
||||||
|
|
||||||
**Nginx 参考配置：**
|
|
||||||
|
|
||||||
```nginx
|
|
||||||
location / {
|
|
||||||
proxy_pass http://127.0.0.1:4000;
|
|
||||||
proxy_buffering off;
|
|
||||||
proxy_cache off;
|
|
||||||
proxy_set_header Connection '';
|
|
||||||
proxy_http_version 1.1;
|
|
||||||
chunked_transfer_encoding off;
|
|
||||||
}
|
|
||||||
```
|
|
||||||
|
|
||||||
</details>
|
|
||||||
|
|
||||||
📖 更详细的接入说明：[docs/client-integration.md](docs/client-integration.md)
|
|
||||||
|
|
||||||
---
|
---
|
||||||
|
|
||||||
|
|||||||
+2
-2
@@ -145,7 +145,7 @@ The AI ecosystem is seeing a growing number of aggregation relay stations based
|
|||||||
**Downstream Clients** (Cursor · Claude Code · Codex · Open WebUI · Cherry Studio, etc.)
|
**Downstream Clients** (Cursor · Claude Code · Codex · Open WebUI · Cherry Studio, etc.)
|
||||||
 ↓  `Authorization: Bearer <PROXY_TOKEN>`
|
 ↓  `Authorization: Bearer <PROXY_TOKEN>`
|
||||||
**Metapi Gateway**
|
**Metapi Gateway**
|
||||||
 • Unified Proxy `/v1/*` — compatible with all OpenAI / Claude endpoints
|
 • Unified `/v1` proxy for core OpenAI / Claude-compatible endpoints (Responses, Chat Completions, Messages, Completions, Embeddings, Images, Models)
|
||||||
 • Smart Routing Engine — weighted selection by cost, balance, and availability; auto-cooldown & retry on failure
|
 • Smart Routing Engine — weighted selection by cost, balance, and availability; auto-cooldown & retry on failure
|
||||||
 • Model Discovery — auto-aggregates all upstream models with zero config
|
 • Model Discovery — auto-aggregates all upstream models with zero config
|
||||||
 • Format Conversion — transparent bidirectional OpenAI ⇄ Claude conversion
|
 • Format Conversion — transparent bidirectional OpenAI ⇄ Claude conversion
|
||||||
@@ -160,7 +160,7 @@ The AI ecosystem is seeing a growing number of aggregation relay stations based
|
|||||||
### Unified Proxy Gateway
|
### Unified Proxy Gateway
|
||||||
|
|
||||||
- Compatible with **OpenAI** and **Claude** downstream formats, works with all mainstream clients
|
- Compatible with **OpenAI** and **Claude** downstream formats, works with all mainstream clients
|
||||||
- Supports Responses / Chat Completions / Messages / Completions (Legacy) / Embeddings / Images / Models — full API coverage
|
- Supports Responses / Chat Completions / Messages / Completions (Legacy) / Embeddings / Images / Models (current core coverage)
|
||||||
- Full SSE streaming support with automatic format conversion (OpenAI <-> Claude)
|
- Full SSE streaming support with automatic format conversion (OpenAI <-> Claude)
|
||||||
|
|
||||||
### Smart Routing Engine
|
### Smart Routing Engine
|
||||||
|
|||||||
@@ -724,6 +724,49 @@ describe('chat proxy stream behavior', () => {
|
|||||||
expect(response.body).toContain('response.completed');
|
expect(response.body).toContain('response.completed');
|
||||||
});
|
});
|
||||||
|
|
||||||
|
it('emits response.failed without synthetic response.completed when upstream stream fails on /v1/responses', async () => {
|
||||||
|
fetchModelPricingCatalogMock.mockResolvedValue({
|
||||||
|
models: [
|
||||||
|
{
|
||||||
|
modelName: 'upstream-gpt',
|
||||||
|
supportedEndpointTypes: ['/v1/chat/completions'],
|
||||||
|
},
|
||||||
|
],
|
||||||
|
groupRatio: {},
|
||||||
|
});
|
||||||
|
|
||||||
|
const encoder = new TextEncoder();
|
||||||
|
const upstreamBody = new ReadableStream<Uint8Array>({
|
||||||
|
start(controller) {
|
||||||
|
controller.enqueue(encoder.encode('data: {"id":"chatcmpl-r3","model":"upstream-gpt","choices":[{"delta":{"role":"assistant"},"finish_reason":null}]}\n\n'));
|
||||||
|
controller.enqueue(encoder.encode('event: error\ndata: {"type":"error","error":{"message":"upstream stream failed","type":"upstream_error"}}\n\n'));
|
||||||
|
controller.enqueue(encoder.encode('data: [DONE]\n\n'));
|
||||||
|
controller.close();
|
||||||
|
},
|
||||||
|
});
|
||||||
|
|
||||||
|
fetchMock.mockResolvedValue(new Response(upstreamBody, {
|
||||||
|
status: 200,
|
||||||
|
headers: { 'content-type': 'text/event-stream; charset=utf-8' },
|
||||||
|
}));
|
||||||
|
|
||||||
|
const response = await app.inject({
|
||||||
|
method: 'POST',
|
||||||
|
url: '/v1/responses',
|
||||||
|
payload: {
|
||||||
|
model: 'gpt-5.2',
|
||||||
|
input: 'hello',
|
||||||
|
stream: true,
|
||||||
|
},
|
||||||
|
});
|
||||||
|
|
||||||
|
expect(response.statusCode).toBe(200);
|
||||||
|
expect(response.body).toContain('response.failed');
|
||||||
|
expect(response.body).toContain('"status":"failed"');
|
||||||
|
expect(response.body).not.toContain('response.completed');
|
||||||
|
expect(response.body).toContain('[DONE]');
|
||||||
|
});
|
||||||
|
|
||||||
it('preserves Responses-specific payload fields and forwards openai headers on /v1/responses', async () => {
|
it('preserves Responses-specific payload fields and forwards openai headers on /v1/responses', async () => {
|
||||||
fetchMock.mockResolvedValue(new Response(JSON.stringify({
|
fetchMock.mockResolvedValue(new Response(JSON.stringify({
|
||||||
id: 'resp_passthrough',
|
id: 'resp_passthrough',
|
||||||
@@ -1666,7 +1709,7 @@ describe('chat proxy stream behavior', () => {
|
|||||||
expect(targetUrl).toContain('/v1/messages');
|
expect(targetUrl).toContain('/v1/messages');
|
||||||
});
|
});
|
||||||
|
|
||||||
it('prefers OpenAI-compatible endpoint when catalog uses generic openai/anthropic labels', async () => {
|
it('prefers Messages endpoint for claude-family models when catalog uses generic openai/anthropic labels', async () => {
|
||||||
fetchModelPricingCatalogMock.mockResolvedValue({
|
fetchModelPricingCatalogMock.mockResolvedValue({
|
||||||
models: [
|
models: [
|
||||||
{
|
{
|
||||||
@@ -1678,16 +1721,12 @@ describe('chat proxy stream behavior', () => {
|
|||||||
});
|
});
|
||||||
|
|
||||||
fetchMock.mockResolvedValue(new Response(JSON.stringify({
|
fetchMock.mockResolvedValue(new Response(JSON.stringify({
|
||||||
id: 'chatcmpl-openai-first',
|
id: 'msg-claude-first',
|
||||||
object: 'chat.completion',
|
type: 'message',
|
||||||
created: 1_706_000_002,
|
|
||||||
model: 'upstream-gpt',
|
model: 'upstream-gpt',
|
||||||
choices: [{
|
content: [{ type: 'text', text: 'hello from messages endpoint' }],
|
||||||
index: 0,
|
stop_reason: 'end_turn',
|
||||||
message: { role: 'assistant', content: 'hello from openai endpoint' },
|
usage: { input_tokens: 5, output_tokens: 3, total_tokens: 8 },
|
||||||
finish_reason: 'stop',
|
|
||||||
}],
|
|
||||||
usage: { prompt_tokens: 5, completion_tokens: 3, total_tokens: 8 },
|
|
||||||
}), {
|
}), {
|
||||||
status: 200,
|
status: 200,
|
||||||
headers: { 'content-type': 'application/json' },
|
headers: { 'content-type': 'application/json' },
|
||||||
@@ -1705,11 +1744,11 @@ describe('chat proxy stream behavior', () => {
|
|||||||
|
|
||||||
expect(response.statusCode).toBe(200);
|
expect(response.statusCode).toBe(200);
|
||||||
const body = response.json();
|
const body = response.json();
|
||||||
expect(body?.choices?.[0]?.message?.content).toContain('hello from openai endpoint');
|
expect(body?.choices?.[0]?.message?.content).toContain('hello from messages endpoint');
|
||||||
|
|
||||||
expect(fetchMock).toHaveBeenCalledTimes(1);
|
expect(fetchMock).toHaveBeenCalledTimes(1);
|
||||||
const [targetUrl] = fetchMock.mock.calls[0] as [string, any];
|
const [targetUrl] = fetchMock.mock.calls[0] as [string, any];
|
||||||
expect(targetUrl).toContain('/v1/chat/completions');
|
expect(targetUrl).toContain('/v1/messages');
|
||||||
});
|
});
|
||||||
|
|
||||||
it('falls back to /v1/messages when catalog only declares openai and chat endpoint fails', async () => {
|
it('falls back to /v1/messages when catalog only declares openai and chat endpoint fails', async () => {
|
||||||
@@ -1750,7 +1789,7 @@ describe('chat proxy stream behavior', () => {
|
|||||||
method: 'POST',
|
method: 'POST',
|
||||||
url: '/v1/chat/completions',
|
url: '/v1/chat/completions',
|
||||||
payload: {
|
payload: {
|
||||||
model: 'claude-haiku-4-5-20251001',
|
model: 'gpt-4o-mini',
|
||||||
stream: false,
|
stream: false,
|
||||||
messages: [{ role: 'user', content: 'hello' }],
|
messages: [{ role: 'user', content: 'hello' }],
|
||||||
},
|
},
|
||||||
@@ -1805,7 +1844,7 @@ describe('chat proxy stream behavior', () => {
|
|||||||
method: 'POST',
|
method: 'POST',
|
||||||
url: '/v1/chat/completions',
|
url: '/v1/chat/completions',
|
||||||
payload: {
|
payload: {
|
||||||
model: 'claude-haiku-4-5-20251001',
|
model: 'gpt-4o-mini',
|
||||||
stream: false,
|
stream: false,
|
||||||
messages: [{ role: 'user', content: 'hello' }],
|
messages: [{ role: 'user', content: 'hello' }],
|
||||||
},
|
},
|
||||||
@@ -1860,7 +1899,7 @@ describe('chat proxy stream behavior', () => {
|
|||||||
method: 'POST',
|
method: 'POST',
|
||||||
url: '/v1/chat/completions',
|
url: '/v1/chat/completions',
|
||||||
payload: {
|
payload: {
|
||||||
model: 'claude-haiku-4-5-20251001',
|
model: 'gpt-4o-mini',
|
||||||
stream: false,
|
stream: false,
|
||||||
messages: [{ role: 'user', content: 'hello' }],
|
messages: [{ role: 'user', content: 'hello' }],
|
||||||
},
|
},
|
||||||
|
|||||||
@@ -170,6 +170,7 @@ async function handleChatProxyRequest(
|
|||||||
},
|
},
|
||||||
modelName,
|
modelName,
|
||||||
downstreamFormat,
|
downstreamFormat,
|
||||||
|
requestedModel,
|
||||||
);
|
);
|
||||||
let startTime = Date.now();
|
let startTime = Date.now();
|
||||||
|
|
||||||
@@ -664,5 +665,7 @@ function logProxy(
|
|||||||
errorMessage: normalizedErrorMessage,
|
errorMessage: normalizedErrorMessage,
|
||||||
retryCount,
|
retryCount,
|
||||||
}).run();
|
}).run();
|
||||||
} catch {}
|
} catch (error) {
|
||||||
|
console.warn('[proxy/chat] failed to write proxy log', error);
|
||||||
|
}
|
||||||
}
|
}
|
||||||
|
|||||||
@@ -188,7 +188,7 @@ export function normalizeStopReason(raw: unknown): string | null {
|
|||||||
return 'tool_calls';
|
return 'tool_calls';
|
||||||
}
|
}
|
||||||
|
|
||||||
return 'stop';
|
return null;
|
||||||
}
|
}
|
||||||
|
|
||||||
export function toClaudeStopReason(finishReason: string | null | undefined): string {
|
export function toClaudeStopReason(finishReason: string | null | undefined): string {
|
||||||
|
|||||||
@@ -257,6 +257,8 @@ function logProxy(
|
|||||||
errorMessage,
|
errorMessage,
|
||||||
retryCount,
|
retryCount,
|
||||||
}).run();
|
}).run();
|
||||||
} catch {}
|
} catch (error) {
|
||||||
|
console.warn('[proxy/completions] failed to write proxy log', error);
|
||||||
|
}
|
||||||
}
|
}
|
||||||
|
|
||||||
|
|||||||
@@ -172,6 +172,8 @@ function logProxy(
|
|||||||
errorMessage,
|
errorMessage,
|
||||||
retryCount,
|
retryCount,
|
||||||
}).run();
|
}).run();
|
||||||
} catch {}
|
} catch (error) {
|
||||||
|
console.warn('[proxy/embeddings] failed to write proxy log', error);
|
||||||
|
}
|
||||||
}
|
}
|
||||||
|
|
||||||
|
|||||||
@@ -142,6 +142,8 @@ function logProxy(
|
|||||||
errorMessage,
|
errorMessage,
|
||||||
retryCount,
|
retryCount,
|
||||||
}).run();
|
}).run();
|
||||||
} catch {}
|
} catch (error) {
|
||||||
|
console.warn('[proxy/images] failed to write proxy log', error);
|
||||||
|
}
|
||||||
}
|
}
|
||||||
|
|
||||||
|
|||||||
@@ -595,7 +595,7 @@ function emitResponsesEvent(
|
|||||||
|
|
||||||
function buildResponseObject(
|
function buildResponseObject(
|
||||||
state: ResponsesStreamState,
|
state: ResponsesStreamState,
|
||||||
status: 'in_progress' | 'completed',
|
status: 'in_progress' | 'completed' | 'failed',
|
||||||
usage?: UsageSummary,
|
usage?: UsageSummary,
|
||||||
): Record<string, unknown> {
|
): Record<string, unknown> {
|
||||||
const response: Record<string, unknown> = {
|
const response: Record<string, unknown> = {
|
||||||
@@ -837,6 +837,44 @@ function completeResponsesStream(
|
|||||||
return events;
|
return events;
|
||||||
}
|
}
|
||||||
|
|
||||||
|
function failResponsesStream(
|
||||||
|
state: ResponsesStreamState,
|
||||||
|
streamContext: { id: string; model: string; created: number },
|
||||||
|
usage: UsageSummary,
|
||||||
|
payload?: unknown,
|
||||||
|
): string[] {
|
||||||
|
if (state.completed) return [];
|
||||||
|
|
||||||
|
const events: string[] = [];
|
||||||
|
events.push(...ensureResponsesStarted(state, streamContext));
|
||||||
|
|
||||||
|
const responsePayload = isRecord(payload) && isRecord((payload as any).response)
|
||||||
|
? ((payload as any).response as Record<string, unknown>)
|
||||||
|
: null;
|
||||||
|
const errorPayload = isRecord(payload) && isRecord((payload as any).error)
|
||||||
|
? ((payload as any).error as Record<string, unknown>)
|
||||||
|
: null;
|
||||||
|
|
||||||
|
if (responsePayload && asTrimmedString(responsePayload.id)) {
|
||||||
|
state.responseId = ensureResponseId(asTrimmedString(responsePayload.id));
|
||||||
|
}
|
||||||
|
if (responsePayload && asTrimmedString(responsePayload.model)) {
|
||||||
|
state.model = asTrimmedString(responsePayload.model);
|
||||||
|
}
|
||||||
|
|
||||||
|
const failedResponse = buildResponseObject(state, 'failed', usage);
|
||||||
|
if (errorPayload) {
|
||||||
|
failedResponse.error = errorPayload;
|
||||||
|
}
|
||||||
|
|
||||||
|
events.push(emitResponsesEvent(state, 'response.failed', {
|
||||||
|
response: failedResponse,
|
||||||
|
}));
|
||||||
|
events.push('data: [DONE]\n\n');
|
||||||
|
state.completed = true;
|
||||||
|
return events;
|
||||||
|
}
|
||||||
|
|
||||||
function serializeConvertedResponsesEvents(input: {
|
function serializeConvertedResponsesEvents(input: {
|
||||||
state: ResponsesStreamState;
|
state: ResponsesStreamState;
|
||||||
streamContext: { id: string; model: string; created: number };
|
streamContext: { id: string; model: string; created: number };
|
||||||
@@ -930,6 +968,7 @@ export async function responsesProxyRoute(app: FastifyInstance) {
|
|||||||
},
|
},
|
||||||
modelName,
|
modelName,
|
||||||
'responses',
|
'responses',
|
||||||
|
requestedModel,
|
||||||
);
|
);
|
||||||
if (endpointCandidates.length === 0) {
|
if (endpointCandidates.length === 0) {
|
||||||
endpointCandidates.push('responses', 'chat', 'messages');
|
endpointCandidates.push('responses', 'chat', 'messages');
|
||||||
@@ -1046,7 +1085,7 @@ export async function responsesProxyRoute(app: FastifyInstance) {
|
|||||||
if (eventBlock.data === '[DONE]') {
|
if (eventBlock.data === '[DONE]') {
|
||||||
if (passthroughResponsesStream) {
|
if (passthroughResponsesStream) {
|
||||||
reply.raw.write('data: [DONE]\n\n');
|
reply.raw.write('data: [DONE]\n\n');
|
||||||
} else {
|
} else if (!responsesState.completed) {
|
||||||
writeLines(completeResponsesStream(responsesState, streamContext, parsedUsage));
|
writeLines(completeResponsesStream(responsesState, streamContext, parsedUsage));
|
||||||
}
|
}
|
||||||
continue;
|
continue;
|
||||||
@@ -1069,6 +1108,20 @@ export async function responsesProxyRoute(app: FastifyInstance) {
|
|||||||
continue;
|
continue;
|
||||||
}
|
}
|
||||||
|
|
||||||
|
const payloadType = (isRecord(parsedPayload) && typeof parsedPayload.type === 'string')
|
||||||
|
? parsedPayload.type
|
||||||
|
: '';
|
||||||
|
const isFailureEvent = (
|
||||||
|
eventBlock.event === 'error'
|
||||||
|
|| eventBlock.event === 'response.failed'
|
||||||
|
|| payloadType === 'error'
|
||||||
|
|| payloadType === 'response.failed'
|
||||||
|
);
|
||||||
|
if (isFailureEvent) {
|
||||||
|
writeLines(failResponsesStream(responsesState, streamContext, parsedUsage, parsedPayload));
|
||||||
|
continue;
|
||||||
|
}
|
||||||
|
|
||||||
if (parsedPayload && typeof parsedPayload === 'object') {
|
if (parsedPayload && typeof parsedPayload === 'object') {
|
||||||
const normalizedEvent = normalizeUpstreamStreamEvent(parsedPayload, streamContext, modelName);
|
const normalizedEvent = normalizeUpstreamStreamEvent(parsedPayload, streamContext, modelName);
|
||||||
writeLines(serializeConvertedResponsesEvents({
|
writeLines(serializeConvertedResponsesEvents({
|
||||||
@@ -1107,7 +1160,7 @@ export async function responsesProxyRoute(app: FastifyInstance) {
|
|||||||
}
|
}
|
||||||
} finally {
|
} finally {
|
||||||
reader.releaseLock();
|
reader.releaseLock();
|
||||||
if (!passthroughResponsesStream) {
|
if (!passthroughResponsesStream && !responsesState.completed) {
|
||||||
writeLines(completeResponsesStream(responsesState, streamContext, parsedUsage));
|
writeLines(completeResponsesStream(responsesState, streamContext, parsedUsage));
|
||||||
}
|
}
|
||||||
reply.raw.end();
|
reply.raw.end();
|
||||||
@@ -1252,5 +1305,7 @@ function logProxy(
|
|||||||
errorMessage: normalizedErrorMessage,
|
errorMessage: normalizedErrorMessage,
|
||||||
retryCount,
|
retryCount,
|
||||||
}).run();
|
}).run();
|
||||||
} catch {}
|
} catch (error) {
|
||||||
|
console.warn('[proxy/responses] failed to write proxy log', error);
|
||||||
|
}
|
||||||
}
|
}
|
||||||
|
|||||||
@@ -1,6 +1,6 @@
|
|||||||
import { beforeEach, describe, expect, it, vi } from 'vitest';
|
import { beforeEach, describe, expect, it, vi } from 'vitest';
|
||||||
|
|
||||||
const fetchModelPricingCatalogMock = vi.fn(async (_arg?: unknown) => null);
|
const fetchModelPricingCatalogMock = vi.fn(async (_arg?: unknown): Promise<any> => null);
|
||||||
|
|
||||||
vi.mock('../../services/modelPricingService.js', () => ({
|
vi.mock('../../services/modelPricingService.js', () => ({
|
||||||
fetchModelPricingCatalog: (arg: unknown) => fetchModelPricingCatalogMock(arg),
|
fetchModelPricingCatalog: (arg: unknown) => fetchModelPricingCatalogMock(arg),
|
||||||
@@ -60,6 +60,41 @@ describe('resolveUpstreamEndpointCandidates', () => {
|
|||||||
expect(responsesOrder).toEqual(['responses', 'chat', 'messages']);
|
expect(responsesOrder).toEqual(['responses', 'chat', 'messages']);
|
||||||
});
|
});
|
||||||
|
|
||||||
|
it('prioritizes messages-first for claude-family models on openai downstream', async () => {
|
||||||
|
fetchModelPricingCatalogMock.mockResolvedValue({
|
||||||
|
models: [
|
||||||
|
{
|
||||||
|
modelName: 'claude-opus-4-6',
|
||||||
|
supportedEndpointTypes: ['anthropic', 'openai'],
|
||||||
|
},
|
||||||
|
],
|
||||||
|
groupRatio: {},
|
||||||
|
});
|
||||||
|
|
||||||
|
const order = await resolveUpstreamEndpointCandidates(
|
||||||
|
{
|
||||||
|
...baseContext,
|
||||||
|
site: { ...baseContext.site, platform: 'new-api' },
|
||||||
|
},
|
||||||
|
'claude-opus-4-6',
|
||||||
|
'openai',
|
||||||
|
);
|
||||||
|
|
||||||
|
expect(order).toEqual(['messages', 'chat', 'responses']);
|
||||||
|
|
||||||
|
const aliasedOrder = await resolveUpstreamEndpointCandidates(
|
||||||
|
{
|
||||||
|
...baseContext,
|
||||||
|
site: { ...baseContext.site, platform: 'new-api' },
|
||||||
|
},
|
||||||
|
'upstream-gpt',
|
||||||
|
'openai',
|
||||||
|
'claude-haiku-4-5-20251001',
|
||||||
|
);
|
||||||
|
|
||||||
|
expect(aliasedOrder).toEqual(['messages', 'chat', 'responses']);
|
||||||
|
});
|
||||||
|
|
||||||
it('keeps explicit platform priority rules', async () => {
|
it('keeps explicit platform priority rules', async () => {
|
||||||
const openaiOrder = await resolveUpstreamEndpointCandidates(
|
const openaiOrder = await resolveUpstreamEndpointCandidates(
|
||||||
{
|
{
|
||||||
|
|||||||
@@ -30,6 +30,12 @@ function normalizePlatformName(platform: unknown): string {
|
|||||||
return asTrimmedString(platform).toLowerCase();
|
return asTrimmedString(platform).toLowerCase();
|
||||||
}
|
}
|
||||||
|
|
||||||
|
function isClaudeFamilyModel(modelName: string): boolean {
|
||||||
|
const normalized = asTrimmedString(modelName).toLowerCase();
|
||||||
|
if (!normalized) return false;
|
||||||
|
return normalized === 'claude' || normalized.startsWith('claude-') || normalized.includes('claude');
|
||||||
|
}
|
||||||
|
|
||||||
function headerValueToString(value: unknown): string | null {
|
function headerValueToString(value: unknown): string | null {
|
||||||
if (typeof value === 'string') {
|
if (typeof value === 'string') {
|
||||||
const trimmed = value.trim();
|
const trimmed = value.trim();
|
||||||
@@ -679,7 +685,11 @@ function normalizeEndpointTypes(value: unknown): UpstreamEndpoint[] {
|
|||||||
return Array.from(normalized);
|
return Array.from(normalized);
|
||||||
}
|
}
|
||||||
|
|
||||||
function preferredEndpointOrder(downstreamFormat: EndpointPreference, sitePlatform?: string): UpstreamEndpoint[] {
|
function preferredEndpointOrder(
|
||||||
|
downstreamFormat: EndpointPreference,
|
||||||
|
sitePlatform?: string,
|
||||||
|
preferMessagesForClaudeModel = false,
|
||||||
|
): UpstreamEndpoint[] {
|
||||||
const platform = normalizePlatformName(sitePlatform);
|
const platform = normalizePlatformName(sitePlatform);
|
||||||
|
|
||||||
if (platform === 'gemini') {
|
if (platform === 'gemini') {
|
||||||
@@ -707,6 +717,11 @@ function preferredEndpointOrder(downstreamFormat: EndpointPreference, sitePlatfo
|
|||||||
return ['messages', 'chat', 'responses'];
|
return ['messages', 'chat', 'responses'];
|
||||||
}
|
}
|
||||||
|
|
||||||
|
if (downstreamFormat === 'openai' && preferMessagesForClaudeModel) {
|
||||||
|
// Claude-family models are most stable with native Messages semantics.
|
||||||
|
return ['messages', 'chat', 'responses'];
|
||||||
|
}
|
||||||
|
|
||||||
return ['chat', 'messages', 'responses'];
|
return ['chat', 'messages', 'responses'];
|
||||||
}
|
}
|
||||||
|
|
||||||
@@ -714,8 +729,13 @@ export async function resolveUpstreamEndpointCandidates(
|
|||||||
context: ChannelContext,
|
context: ChannelContext,
|
||||||
modelName: string,
|
modelName: string,
|
||||||
downstreamFormat: EndpointPreference,
|
downstreamFormat: EndpointPreference,
|
||||||
|
requestedModelHint?: string,
|
||||||
): Promise<UpstreamEndpoint[]> {
|
): Promise<UpstreamEndpoint[]> {
|
||||||
const sitePlatform = normalizePlatformName(context.site.platform);
|
const sitePlatform = normalizePlatformName(context.site.platform);
|
||||||
|
const preferMessagesForClaudeModel = (
|
||||||
|
isClaudeFamilyModel(modelName)
|
||||||
|
|| isClaudeFamilyModel(asTrimmedString(requestedModelHint))
|
||||||
|
);
|
||||||
if (sitePlatform === 'anyrouter') {
|
if (sitePlatform === 'anyrouter') {
|
||||||
// anyrouter deployments are effectively anthropic-protocol first.
|
// anyrouter deployments are effectively anthropic-protocol first.
|
||||||
return downstreamFormat === 'responses'
|
return downstreamFormat === 'responses'
|
||||||
@@ -723,7 +743,17 @@ export async function resolveUpstreamEndpointCandidates(
|
|||||||
: ['messages', 'chat'];
|
: ['messages', 'chat'];
|
||||||
}
|
}
|
||||||
|
|
||||||
const preferred = preferredEndpointOrder(downstreamFormat, context.site.platform);
|
const preferred = preferredEndpointOrder(
|
||||||
|
downstreamFormat,
|
||||||
|
context.site.platform,
|
||||||
|
preferMessagesForClaudeModel,
|
||||||
|
);
|
||||||
|
const forceMessagesFirstForClaudeModel = (
|
||||||
|
downstreamFormat === 'openai'
|
||||||
|
&& preferMessagesForClaudeModel
|
||||||
|
&& sitePlatform !== 'openai'
|
||||||
|
&& sitePlatform !== 'gemini'
|
||||||
|
);
|
||||||
|
|
||||||
try {
|
try {
|
||||||
const catalog = await fetchModelPricingCatalog({
|
const catalog = await fetchModelPricingCatalog({
|
||||||
@@ -752,6 +782,25 @@ export async function resolveUpstreamEndpointCandidates(
|
|||||||
if (!matched) return preferred;
|
if (!matched) return preferred;
|
||||||
|
|
||||||
const supportedRaw = Array.isArray(matched.supportedEndpointTypes) ? matched.supportedEndpointTypes : [];
|
const supportedRaw = Array.isArray(matched.supportedEndpointTypes) ? matched.supportedEndpointTypes : [];
|
||||||
|
const normalizedSupportedRaw = supportedRaw
|
||||||
|
.map((item) => asTrimmedString(item).toLowerCase())
|
||||||
|
.filter((item) => item.length > 0);
|
||||||
|
const hasConcreteEndpointHint = normalizedSupportedRaw.some((raw) => (
|
||||||
|
raw.includes('/v1/messages')
|
||||||
|
|| raw.includes('/v1/chat/completions')
|
||||||
|
|| raw.includes('/v1/responses')
|
||||||
|
|| raw === 'messages'
|
||||||
|
|| raw === 'chat'
|
||||||
|
|| raw === 'chat_completions'
|
||||||
|
|| raw === 'completions'
|
||||||
|
|| raw === 'responses'
|
||||||
|
));
|
||||||
|
if (forceMessagesFirstForClaudeModel && !hasConcreteEndpointHint) {
|
||||||
|
// Generic labels like openai/anthropic are too coarse for Claude models;
|
||||||
|
// keep messages-first order in this case.
|
||||||
|
return preferred;
|
||||||
|
}
|
||||||
|
|
||||||
const supported = new Set<UpstreamEndpoint>();
|
const supported = new Set<UpstreamEndpoint>();
|
||||||
for (const endpoint of supportedRaw) {
|
for (const endpoint of supportedRaw) {
|
||||||
const normalizedList = normalizeEndpointTypes(endpoint);
|
const normalizedList = normalizeEndpointTypes(endpoint);
|
||||||
|
|||||||
+1
-1
@@ -6,7 +6,7 @@ import './index.css';
|
|||||||
|
|
||||||
ReactDOM.createRoot(document.getElementById('root')!).render(
|
ReactDOM.createRoot(document.getElementById('root')!).render(
|
||||||
<React.StrictMode>
|
<React.StrictMode>
|
||||||
<BrowserRouter>
|
<BrowserRouter future={{ v7_startTransition: true, v7_relativeSplatPath: true }}>
|
||||||
<App />
|
<App />
|
||||||
</BrowserRouter>
|
</BrowserRouter>
|
||||||
</React.StrictMode>
|
</React.StrictMode>
|
||||||
|
|||||||
Reference in New Issue
Block a user